## Bayesian learning

The predicted probability, P(C), for some future event, C – someone buys your product or service, some student drops out of high school, there is a flood or wildfire, there is an oil pipeline leak, someone gets a disease, and many others – is at the heart of answering the most fundamental questions – Who? What? Where? and When? – for your business or organisation, or as an individual.

The answers to these questions inevitably depend on a host of factors, **X**, that can change over time as new information becomes available. In the light of this information our *prior *estimate P(C) for the event can change P(C) —> P(C|**X**) with P(C|**X**) being our estimate *posterior* to the information **X**. Sometimes we have enough information to calculate the probabilities as frequencies, as when we flip a coin and count how many times we get heads. Sometimes this isn’t possible. When we ask “will it rain tomorrow?” There is only one tomorrow. In this case, the probabilities becomes beliefs or plausibilities. This is the Bayesian interpretation of probability. It is how our brain works and is at the heart of Human Intelligence encapsulating our intuition, expertise and experience.

Our beliefs P(C) can be changed by data **X**. This is learning. It can even be stated mathematically:

P(C|**X**) = P(**X**|C)P(C)/P(**X**)

where P(**X**|C) is the *likelihood* that we observe the data **X **in the cases where the event C occurs. Although we can learn in the light of new data, often that learning is inadequate. In fact, this is at the heart of many of humanity’s greatest problems, where, unfortunately, we choose that data that best matches our beliefs.

It is the Data Revolution that can provide information that allows us to potentially correct the biases in our beliefs. Artificial Intelligence is then needed to calculate P(**X**|C) when many factors **X** are involved. At the heart of **Presage’s** HIPRE Platform is this combination of Human and Artificial Intelligence using a Bayesian framework to create prediction models that are transparent so that the contributions of each individual factor to P(C|**X**) can be identified, quantified, interpreted using Human Intelligence, and then acted upon.